Saturday, April 9, 2011

Whose house?

I occurred to me that I'm sort of competing with Mark Cave-Ayland. While Mark makes a really good job to get a proprietary OS (Solaris 8) working with an Open Source firmware (OpenBIOS) under a 32 bit SPARC qemu machine, I'm trying to get a Open Source OS (OpenSolaris) working with a proprietary firmware (OBP) under a 64 bit SPARC qemu machine. Right now we are at about the same place - he is right after the disk detection, I'm a bit behind.

Who's gonna see the '#' first, what are your bets? What task do you think is easier?

But speaking seriously, if Mark makes it first I'll buy him a beer.


Anonymous said...

I hope to see early a working qemu-spac64 system.

build and test a program under both x86 and spac64 is a good certificate for a portable bugfree codebase.

Go Artyom Go :-)

atar said...

But does it have to be a sparc64? To test the portability a sparc32 would probably be a good starter, right?

Jason Stevens said...

I would imagine that running OBP you'd get better luck running Solaris (any flavor) since it is far closer to the real deal...

Good luck Artyom!

atar said...

True, but to make it run this way a lot of additional devices have to be implemented. They are sort of documented, but in reality a lot of edge cases is not covered.

For instance the interrupt processing is level triggered in both CPU and IOMMU, and there are registers to acknowledge and disable the interrupts. The normal workflow is well defined. But what happens if the IRQ is lowered before it gets acknowledged? Does disabling irq acknowledge them? And so on...

On the other hand OpenBIOS perfectly lives without IOMMU.